The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data in the form of legal contracts provided via text, files, or URLs. It lacks explicit sanitization or boundary markers to prevent the agent from following instructions that might be embedded within the contract text itself.
Ingestion points: untrusted data enters the context via the Read tool from files and URLs as specified in the Step 1 workflow of SKILL.md.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are provided to the model regarding the contract content.
Capability inventory: The skill has access to Bash, Write, Read, Glob, and Grep tools to perform its tasks.
Sanitization: The instructions do not define any sanitization, escaping, or validation logic for the external content before it is processed by the model.
[EXTERNAL_DOWNLOADS]: The workflow in SKILL.md explicitly instructs the agent to ingest contract content from a URL. This involves making network requests to arbitrary external domains provided by the user, which could be used to fetch malicious content or interact with untrusted servers.

contract-redliner